A Survey of the Yolo Series of Object Detection Algorithms
Jiahao Xu, Jianping Li, Zhen Zhou, Qing Lv, Jiaheng Luo
Abstract
This paper comprehensively reviews the YOLO series algorithms in computer vision. It begins by emphasizing the significance of object detection in security, driving and image analysis, and then traces the evolution of the YOLO series from v1. Detailed introductions are given to each version's principles including network design, detection methods and loss functions, along with iterative improvements. Their performance is evaluated on standard datasets and influencing factors are analyzed. The application scenarios cover security, driving and industry, confirming their practicality. Nevertheless, challenges like detecting small and occluded objects, poor interpretability and high resource demands remain. Finally, the paper summarizes YOLO's contributions and looks ahead to future research on performance improvement, interpretability enhancement and technology integration, providing valuable insights for object detection research and application.